2

I am student in biology. For my examination I conducted a greenhouse experiment with gras. Actually I am not sure how to develop the correct model and unfortunately I haven't found an answer in the web during the last weeks.

Experiment:

4 different species of gras were planted in pots with 4 different types of fertilizers (different N content).

After the first harvest fertilizers were changed until second harvest. The experiment was completely randomized and consisted of six repetitions.

Question:

I am looking for an appropriate model and a post-hoc test. I guess modeling each harvest by its own using typical anova is no problem. For analyzing sum of biomass I favor repeated measures models.

Model:
Columns:
Pot_ID: ID number of the pot

Species: Name of the species
Fertilizer: A, B, C, D

Time: 1 for first harvest, 2 for total biomass including second harvest

Yield: Biomassproduction

model=lme(yield~species*fertilizer,random=~1|pot_ID,dataset,correlation=corAR1(form=~time|pot_ID))

Is the model appropriate? What does anova(model) show, both times or only time=2? How to implement a post-hoc?

Manuel Saglund
  • 21
  • 1

1 Answers1

0

I think modeling the first harvest using ANOVA is not a problem; it gets quite complicated for the second harvest because prior effects will be very different depending on the residual effects of the previous N content (left in the soil and accumulated by the plant), which is unlikely to be a random effect or be completely accounted for by the first harvest biomass. A complete design accounting for that would have 24 (4! or factorial of all N trt) treatment combinations (each prior fertilizer matched to each type of fertilizer in the second round). You could try to rephrase the question, looking at the rate of biomass increase depending on the sign of change in the fertilizer (increase or decrease).

This has a nice detailed discussion of post-hoc comparisons How to obtain the results of a Tukey HSD post-hoc test in a table showing grouped pairs?

Community
  • 1
katya
  • 2,084
  • 8
  • 11
  • Thank you very much. I agree with you. Reprasing might be a good idea. I can seperate this experiment into two experiments. First one until the first harvest. Second one starting with plants that were grown under different nutrient conditions (until first cut). So I avoid questions of repeated measures. – Manuel Saglund Nov 04 '14 at 12:54